By Eric Léopold, Founder, Threedot
In 1995 I was programming algorithms based on neural networks, with a dream that one day we would achieve a breakthrough in computer science. 30 years later, it is taking shape as we are about to solve one of humanity’s biggest headache: booking a flight😉
Two years ago, OpenAI released a conversational bot called ChatGPT, powered by Large Language Models (LLM) running on neural networks. These models created the category called Generative AI, or GenAI, as it can generate text that looks like human-generated text.
The latest generation of artificial intelligence (AI) algorithms is called “AI Agent”, whereby an agent is a system capable of autonomous decision-making towards achieving a goal. To understand how these AI agents could reshape travel booking, let’s explore what makes these systems unique and what role they can play.
NB: let’s not confuse AI Agents with the real Air India travel agents 😉
The attributes of an AI agent
An AI agent relies on decision-making and interaction capabilities, based on perception and memory, that support its autonomy to achieve a goal.
Current systems are rule-based, meaning that a programmer needs to think through all the situations and programs all the rules in the system. Once the system is programmed, it becomes autonomous, like an aircraft on autopilot, interacting with its environment but without human intervention.
The biggest challenge of the AI agent is the decision-making capability towards a goal. In the rule-based model, the decisions are hard-coded in the rules by humans, whereas the AI Agent needs to perform a kind of live reasoning to make a decision. For example, the system will realize it faces a challenge, will assess all the options and decide on a course of action that minimize the effort to achieve a goal.
The promise of the AI agent is indeed to learn the rules, in what is usually called “machine learning”, rather than follow them as coded. For example, a self-driving car can learn how to drive autonomously, without human intervention, by recording situations from a human driver. In the case of a car, the perception of the environment is critical to the success of the experience. AI collects inputs from video feed and sensors to build a model of the environment surrounding the car.
An AI agent, that has learnt a language or to drive a car, needs to memorize a specific context related to a mission. The memory lasts beyond a session and covers multiple sources relevant to the domain of expertise of the agent.
Finally, an AI agent must be able to interact with the environment and trigger actions. These actions require a deep integration with all digital systems. The AI agent will act on behalf of the human being that gave it a mission to fulfil. The delegation of authority may include credentials and other resources. In the physical world, AI agents will be embedded in humanoids that can use their physical force to achieve a goal.
How AI agents will transform travel booking
What does the rise of AI agents mean for travel agents and the rest of the travel value chain? Will it change the way consumers and professionals shop for flights? Will airlines have to rethink the way they distribute their products?
From a consumer perspective, shopping for flights changed with the emergence of the web in the 1990s. But in the past two decades, web shopping for flights has not changed a lot: enter “origin airport, destination airport and date”, review a long list of options and combinations, and pick one itinerary.
From a travel agent perspective, the web enabled a new category to emerge, the online travel agents or OTAs, and a parallel category, that is not an agent in the sense of selling tickets, that enables the search for flights, the meta-search websites. Like for consumers, there was no major changes in recent years.
What may change with the development of AI agents is the full delegation and automation of travel search. While the complexity of travel itineraries, combining multiple airlines (interline and codeshares), modes of transport (intermodal) and connectivity (airport hubs), may not go away, the user interface to navigate them and find the most relevant is about to evolve significantly.
The AI agent will memorize the customer’s personal details and preferences, including payment, and apply them to a travel objective, from a simple return flight to a comprehensive holiday package. The discovery phase will expand beyond the flights and hotels to the experiences and other potential activities that will make the journey really unique. The operational phase will assist the traveler during the entire journey.
The future of human interaction in the age of AI
Executive assistants and travel assistants add a lot of value. The extent to which they will embrace technology to become more effective faster than consumers will embrace technology and become assisted by AI agents is to be determined in the coming years.
The real opportunity is to make travel booking fully autonomous, not just having people talk to bots that replicate the existing complexity. Success for an AI agent is to anticipate a leisure or business trip and make a full packaged proposal, without even asking for it.
In any case, human beings will mostly favor human interactions. Like the web and video calls increased the number of human interactions, rather than replaced them, I believe that AI agents will increase productivity and human interactions overall.
For more articles by Léopold see:
- What does airlines’ New Payment Capability (NPC) look like?
- The dual mode of airline innovation in technology
- Five opportunities to improve travelers’ experience post-COVID
